The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze a 'writing corpus' provided by the user, which serves as an entry point for untrusted external data. Ingestion points: The 'Writing corpus' described in SKILL.md is the primary source of external content. Boundary markers: The skill instructions provide 'Corpus Quality Gates' to filter noise and artifacts but lack explicit security delimiters or instructions to ignore potential adversarial commands embedded within the corpus. Capability inventory: The agent's capabilities in this skill are restricted to text analysis and the generation or refinement of Markdown documents; it does not invoke tools for network access, sensitive file reads, or shell command execution. Sanitization: The distillation workflow relies on identifying recurring patterns and quoting direct evidence, which grounds the output in the source text but does not include explicit sanitization for indirect prompt injection.

voice-distiller